Animal Kingdom Dataset

Instructions for using the Animal Kingdom Dataset

Reading README files

The README files can be opened using any text editor. Alternatively, rename .md to .txt to view the contents.

Dataset and Code

Download the dataset and code from https://sutdcv.github.io/Animal-Kingdom

File size of dataset (~80GB in total)

• action_recognition/dataset/video.tar.gz
  • 15.6GB –> 15.7GB uncompressed
• action_recognition/dataset/image.tar.gz
  • 42.1GB –> 43.5GB uncompressed
• action_recognition/annotation
  • 271.6 MB
• pose_estimation/dataset/dataset.tar.gz
  • 2.3GB –> 2.3GB uncompressed
• pose_estimation/annotation
  • 259.3 MB
• video_grounding/dataset.tar.gz
  • 15.1GB –> 15.2GB uncompressed
• video_grounding/features.tar.gz
  • 1.1GB –> 1.3GB uncompressed
• video_grounding/annotation
  • 1.1 MB

Structure of folder

• Read the respective README_$TASK.MD files to set up the respective environment to run the baseline codes
• The respective codes are obtained from the original source.
  • Modifications of respective codes can be found in respective $TASK/code_new folders
• Helper scripts ({task}/code/code_new/prepare_dir_{task}.sh) are provided to automatically set up the environment to directly run our dataset. Please read the respective {task}/README_{task}.md files